Mass exponent reports - CSVs

[Madpoo]

I had a request from someone lately for some data that they had been getting by crawling the site, but this seemed like a much more efficient way.

First up as a dump of the Primenet "factoring effort" data.

It has the exponent, the trial-factor bit level, and the B1/B2 bounds (this table, by the way, does not include exponents with a known factor).

Factoring Effort - 2019-03-02

Note that it's a large 7z file (27MB) that decompresses to a quite large CSV.

Coming up, he had asked for some sort of dump of the LL work on each exponent (whether it was unverified/verified as in single/double checked with a match).

When I have that ready I'll link to it here too.
EDIT: Here it is... all exponents with their status (verified/unverified/suspect) and their residue (masked for unverified/suspect). I didn't include proven bad results or any exponent where a factor was found.
LL Status - 2019-03-02

The goal being that if you have some need for data and are having to rely on crawling the exponent report page, it's terribly inefficient and if it's a quick query into a CSV, I can put something or another together.

PS - No guarantees on how long these files may be reachable... they're "point in time" so it may disappear in a few months or a year when it's suitably stale and I feel like cleaning up old files.

[nomead]

Thanks for the list. Although I'm not the one who requested it, I have to admit browsing through the exponent data from time to time. Here and there, there are interesting "peaks" in the factoring effort and this makes it a whole lot easier to find them. Not any valid mathematical or scientific reason, just out of interest.

But another request. Would it also be possible to get a list of found factors? I can do all the sorting and filtering from there. The purpose is to make a list of known factors for various bit levels and exponent sizes, both "easy" and "hard" to find, but a larger list than the current built-in self test list included in mfaktc. Just to be more certain that any of the poking around I may be doing in the software won't result in missed factors.

[Madpoo]

Originally Posted by nomead

Thanks for the list. Although I'm not the one who requested it, I have to admit browsing through the exponent data from time to time. Here and there, there are interesting "peaks" in the factoring effort and this makes it a whole lot easier to find them. Not any valid mathematical or scientific reason, just out of interest.

But another request. Would it also be possible to get a list of found factors? I can do all the sorting and filtering from there. The purpose is to make a list of known factors for various bit levels and exponent sizes, both "easy" and "hard" to find, but a larger list than the current built-in self test list included in mfaktc. Just to be more certain that any of the poking around I may be doing in the software won't result in missed factors.

James has downloadable lists of the factor data over at mersenne.ca

I don't remember where those are but I'd recommend PM'ing him for details - they're quite large and he may not want everyone piling on and snagging those willy-nilly.

[nomead]

Originally Posted by Madpoo

James has downloadable lists of the factor data over at mersenne.ca

I don't remember where those are but I'd recommend PM'ing him for details - they're quite large and he may not want everyone piling on and snagging those willy-nilly.

I found one package on download.mersenne.ca but that's from 2008-02-04 and only about 8 MB. But I'll PM him, let's see. Maybe it's possible to get the >1000M data as well, that would be even better.

[Madpoo]

Originally Posted by nomead

I found one package on download.mersenne.ca but that's from 2008-02-04 and only about 8 MB. But I'll PM him, let's see. Maybe it's possible to get the >1000M data as well, that would be even better.

Correct... if you want factors of exponents over 1e9 then James is the man.

[Yxine]

Madpoo, thanks a lot!

Some questions:
1. In llt report there are letters U, V and S. U = unverified, V = verified, S = suspect (different resudes). Is there any other letters or it is a full list?
2. Will this reports be regullary updated?

Nomead, I can generate for you the list with factors. Will send you a link soon here.

[Madpoo]

Originally Posted by Yxine

Madpoo, thanks a lot!

Some questions:
1. In llt report there are letters U, V and S. U = unverified, V = verified, S = suspect (different resudes). Is there any other letters or it is a full list?
2. Will this reports be regullary updated?

Nomead, I can generate for you the list with factors. Will send you a link soon here.

Yes, that's correct:
S = Suspect
U = Unverified
V = Verified

On average, a "suspect" result has a 50/50 chance of being good. That's why the server will re-assign an exponent for a new *first* time check if the previous result was flagged as suspect based on the error code.

Currently there are 884 exponents with a pair of mismatched tests (672 of those are available to be assigned, FYI)... you'll see those entries listed twice in the data with different masked residues.

[Yxine]

Madpoo, will these reports be updated regulary?

[James Heinrich]

Originally Posted by Madpoo

James has downloadable lists of the factor data over at mersenne.ca
I don't remember where those are but I'd recommend PM'ing him for details - they're quite large and he may not want everyone piling on and snagging those willy-nilly.

All the data exports from mersenne.ca live at https://www.mersenne.ca/export/
The archives are updated every Sunday morning around 0200h UTC.
Feel free to download whatever you want. The original poster is probably most interested in the section at the bottom of the table, notably the "TF Limit" and "Factors n,k" rows.

[kriesel]

Originally Posted by Madpoo

I had a request from someone lately for some data that they had been getting by crawling the site, but this seemed like a much more efficient way.

First up as a dump of the Primenet "factoring effort" data.

It has the exponent, the trial-factor bit level, and the B1/B2 bounds (this table, by the way, does not include exponents with a known factor).

Factoring Effort - 2019-03-02

Note that it's a large 7z file (27MB) that decompresses to a quite large CSV.

Coming up, he had asked for some sort of dump of the LL work on each exponent (whether it was unverified/verified as in single/double checked with a match).

When I have that ready I'll link to it here too.
EDIT: Here it is... all exponents with their status (verified/unverified/suspect) and their residue (masked for unverified/suspect). I didn't include proven bad results or any exponent where a factor was found.
LL Status - 2019-03-02

The goal being that if you have some need for data and are having to rely on crawling the exponent report page, it's terribly inefficient and if it's a quick query into a CSV, I can put something or another together.

PS - No guarantees on how long these files may be reachable... they're "point in time" so it may disappear in a few months or a year when it's suitably stale and I feel like cleaning up old files.

So where does one go for PRP status data?

[Madpoo]

Originally Posted by kriesel

So where does one go for PRP status data?

I just created new "archives of interesting things". I included the archive file sizes so you know what you're getting in to. The archive of the factors is large, and decompresses to 1.05 GB.

Known factors (for exponents < 1e9 of course):
347 MB
https://www.mersenne.org/result_arch..._2019-05-13.7z

Factoring effort:
27 MB
https://www.mersenne.org/result_arch..._2019-05-13.7z

LL status:
18 MB
https://www.mersenne.org/result_arch..._2019-05-13.7z

PRP status:
< 1 MB
https://www.mersenne.org/result_arch..._2019-05-13.7z

Note that the PRP status includes an "R" status type which indicates it was done with the Reliable testing method. It also includes the residue base and type since that matters.